8047 2020

January 2020

Where Does Multinational Profit Go with Territorial Taxation? Evidence from the UK Dominika Langenmayr, Li Liu

Impressum:

CESifo Working Papers ISSN 2364-1428 (electronic version) Publisher and distributor: Munich Society for the Promotion of Economic Research - CESifo GmbH The international platform of Ludwigs-Maximilians University’s Center for Economic Studies and the ifo Institute Poschingerstr. 5, 81679 Munich, Germany Telephone +49 (0)89 2180-2740, Telefax +49 (0)89 2180-17845, email office@cesifo.de Editor: Clemens Fuest www.cesifo-group.org/wp

An electronic version of the paper may be downloaded · from the SSRN website: www.SSRN.com · from the RePEc website: www.RePEc.org · from the CESifo website: www.CESifo-group.org/wp

CESifo Working Paper No. 8047

Where Does Multinational Profit Go with Territorial

Taxation? Evidence from the UK

Abstract In 2009, the United Kingdom abolished the taxation of profits earned abroad and introduced a territorial tax system. Under the territorial system, firms have strong incentives to shift profits abroad. Using a difference-in-differences research design, we show that profits of UK subsidiaries in low-tax countries increased after the reform compared to subsidiaries of non-UK multinationals in the same countries, by an average of 2.1 percentage points. The increase in profit shifting also leads to increases in measured productivity of the foreign affiliates of UK multinationals of between 5 and 9 percent.

JEL-Codes: H250, H870, F230.

Keywords: profit shifting, territorial tax system, multinational firms.

Dominika Langenmayr KU Eichstätt-Ingolstadt

Ingolstadt School of Management Auf der Schanz 49

Germany – 85049 Ingolstadt dominika.langenmayr@ku.de

Li Liu International Monetary Fund 1900 Pennsylvania Avenue

USA – Washington DC, 20009 lliu@imf.org

10th January 2020 We are grateful to John Damstra and Dongxian Guo for providing excellent research assistance. We would like to thank Kimberly Clausing, Daniel Dias, Jim Hines, Michael Keen, Martin Simmler, Michael Smart, and seminar participants at the IMF, Oxford, the National Tax Association 2018 and the IIPF 2019 meetings for helpful comments. Any remaining errors are our own. Work on this project was started when Dominika Langenmayr visited the Oxford University Centre for Business Taxation, she is grateful for their hospitality. She also thanks the German Research Foundation for funding (grant LA 3565-4/1). The views expressed are the authors’ and do not necessarily represent the views of the IMF, its Executive Board, or IMF management.

1 Introduction

The U.S. tax reform of 2017 – known as the Tax Cuts and Jobs Act (TCJA) – may well prove

a landmark event for the international tax system. Beyond the reduction in the statutory

federal tax rate, from 35 to 21 percent, and a move towards full expensing of domestic

investments, it also introduced a territorial tax system exempting U.S. multinationals’ active

business income earned abroad from being taxed in the U.S. upon repatriation. Instead,

foreign profits are taxed only once in the country where the affiliate is active. The move

towards territoriality comes with an important qualification: the TCJA imposes a minimum

tax on the overseas income that exceeds a 10 percent return on tangible assets. The U.S.

introduced this additional tax to address the concern that territorial taxation is associated

with greater incentives for outward profit shifting.

We provide evidence that this concern was warranted by studying the introduction of a

territorial tax system in the United Kingdom in 2009. Our paper thus contributes to the small

literature studying empirically the up- and downsides of different international corporate tax

systems. How to tax foreign profits of multinationals is one of the most fundamental aspects

of corporate taxation worldwide, especially with increasing prominence of multinationals in

the global market. Despite its importance, only few empirical studies have considered this

question. In particular, no previous paper analyzes how the introduction of a territorial tax

system affects profit shifting and firm productivity.1 This question is of particular importance

as the incentives to shift profits abroad are stronger under a territorial system and may cost

government significant tax revenues without effective safeguards in place. Thus, an increase

in profit shifting is often considered one of the main disadvantages of introducing a territorial

tax system.

To answer this question, we use a difference-in-differences (DID) research design com-

paring affiliates of UK multinationals with affiliates of multinational firms headquartered in

other countries. If these affiliates are active in a country with a lower tax rate than the UK,

we expect that UK multinationals shifted more profits to their subsidiaries after 2009. Our

findings confirm this hypothesis, indicating that profitability (measured by earnings before

1In a concurrent working paper, Liu et al. (2019) study a specific aspect of profit shifting, namely transfermispricing.

2

tax relative to total assets) increase by 2.1 percentage points, or an eleven percent increase

in their pre-tax earnings. This is a large effect, as the extent of increase in pre-tax profits is

equivalent to a seven percentage point reduction in the statutory corporate income tax rate

in the UK. Further tests show that firms with high levels of intangible assets are reacting

especially strongly. To confirm that indeed profit shifting (and not increases in investment

or sales) are driving the results, we conduct placebo tests with affiliates in countries that

have higher tax rates than the UK (and where firms thus have no incentive to shift profits),

and find no effect.

Furthermore, we use consolidated accounts to study the global profitability of UK mul-

tinationals. In this setting, we use domestic companies in the UK as the control group.

We find no effect of the reform on the overall profitability of UK multinationals, indicating

that the effect found on the affiliate level arises because of profit shifting and not because of

higher profitability of UK multinationals in general.

Profit shifting not only affects the distribution of tax revenue across countries, but can

also distort the measurement of economic indicators such as firm-level productivity: for ex-

ample, the manipulation of transfer prices would inflate reported productivity as it artificially

increases turnover in low-tax countries. To gauge the importance of this mismeasurement,

we study how total factor productivity (TFP) of the affiliates of UK multinationals develops

after the territorial tax reform. In low-tax countries, measured TFP increases by 5-9%. In

high-tax countries, we find precisely estimated insignificant coefficients, indicating that there

was no effect in measured productivity among UK affiliates in high-tax countries. Thus, the

change in the low-tax countries likely results from profit shifting, and does not reflect real

changes in productivity.

Our results have important implications for policymakers. For the reforming country,

they highlight that measures against profit shifting should accompany the introduction of a

territorial tax system. Indeed, U.S. politicians cited concerns about outward profit shifting

under a territorial tax system as the justification for the various anti-avoidance measures

that accompany the territorial tax reform in the Tax Cuts and Jobs Act. At the same time,

the U.S. tax reform also decreased the federal corporate tax rate substantially, from 35% to

21%, which should also decrease the incentives to shift profits abroad. For countries that

3

aim to use their tax system to attract both real investment and paper profits, our results

highlight the potentially intensified pressure on lowering their statutory tax rates. On a

global scale, as the territorial tax system has become more and more common over time, the

empirical evidence also supports introducing a minimum tax on outbound investments of

multinational firms to alleviate profit shifting and tax competition, for example in a similar

vein as the Global Anti-Base Erosion (GloBE) tax that has recently been proposed by the

OECD and aims to ensure a minimum level of taxation of multinationals globally.2

Our paper contributes to several lines of literature on taxing multinational firms. First,

we add to the literature that compares the territorial and worldwide tax systems in terms

of economic efficiency. In a groundbreaking theoretical analysis, Peggy Musgrave (nee Rich-

man) suggested that worldwide taxation, when implemented in its pure form, achieves “cap-

ital export neutrality” (CEN): As the tax burden is independent of the location of earnings,

it does not distort the allocation of capital across jurisdictions (Richman, 1963). However,

CEN is not robust to deferral and inversion; and it is never fully achieved because there is

no refund when foreign tax credits exceed home tax liabilities. Territoriality, by contrast,

provides for “capital import neutrality” (CIN), meaning that all investments in a particular

country face the same tax rate, regardless of their source. Desai and Hines (2003, 2004)

have pointed out that a territorial tax system is optimal if foreign production takes place

in addition to (and not instead of) domestic production.3 Becker and Fuest (2011) compare

these arguments and highlight that higher compliance costs in a worldwide tax system may

also pose an argument in favor of territorial taxation. More recently, Devereux et al. (2015)

synthesize and extend the literature, showing that a cash-flow tax and a worldwide tax sys-

tem ensure both the optimal allocation of mobile factors and efficient foreign investment. In

practice, however, it is impossible to ensure all dimensions off tax neutrality (e.g. CEN, CIN

and CON) without fully harmonized tax systems (IMF, 2019).

Empirical studies have analyzed the impact of a change in the international tax regime

2For more details on the GloBE, see for example OECD, 2019, “Programme of Work to Develop aConsensus Solution to the Tax Challenges Arising from the Digitalisation of the Economy”, OECD/G20Inclusive Framework on BEPS, OECD, Paris.

3“Capital ownership neutrality” (CON) is another important concept, meaning that taxes do not distortthe ownership of capital. It is achieved either by global adoption of a territorial system or of a worldwidesystem with tax credits.

4

on various margins of firm behavior, including on dividend repatriation (Egger et al., 2015;

Hasegawa and Kiyota, 2017), cash holding/dividend payout (Arena and Kutner, 2015), real

investment (Liu, 2019), and mergers and acquisitions (Feld et al., 2016).4 Focusing on

international transfer mispricing as a particular channel of profit shifting, Liu et al. (2019)

find that the UK territorial tax reform reduced transfer prices on UK exports to low-tax

countries by more than one third. Azemar and Dharmapala (2019) show that introducing

a territorial tax system does not affect the value of tax sparring agreements (which prevent

host country tax incentives from being nullified by home country taxes).5

We also contribute to the more general literature on profit shifting by multinational

companies, recently surveyed by Dharmapala (2014), Heckemeyer and Overesch (2017) and

Beer et al. (2019). Studies have provided evidence for profit shifting using tax rate changes

(e.g. Huizinga and Laeven, 2008; Weichenrieder, 2009; Buettner and Wamser, 2013) or firm-

level earnings shocks (Dharmapala and Riedel, 2013) for identification. We use the change

in profit shifting incentives that follow from abolishing the worldwide tax system.6 Clausing

(2003, 2009) and Desai et al. (2006) provide evidence for profit shifting and deferral in

countries with worldwide tax systems.7 This paper also joins a growing empirical literature

that examines the real effects of international tax avoidance on multinationals’ investment

and employment (Overesch, 2009; de Mooij and Liu, 2020; Suarez Serrato, 2018), and on

country-level aggregate productivity growth (Guvenen et al., 2017).

4Before the 2009 tax regime changes in the UK and Japan, several studies attempted to infer the effectsof the worldwide vs. territorial tax systems by comparing firm behavior in countries with a worldwide taxsystem with that in countries with a territorial tax system. Maffini (2012) and Markle (2016) show thatthere is less tax avoidance in a worldwide system. Differences between multinationals domiciled in territorialand worldwide countries have also been found along other dimensions, such as the location of foreign directinvestment (Hines, 1996; Clausing and Shaviro, 2011), headquarter relocations (Voget, 2011), and subsidiarylocation choices (Barrios et al., 2012).

5There is also a literature studying tax-repatriation holidays, which are similar to a temporary switchto a territorial tax system, see e.g. Dharmapala et al. (2011) on domestic investment and Flaaen (2017) ontransfer mispricing.

6Multinational companies in countries with a worldwide tax system often postpone repatriation of foreignearnings to avoid paying home-country corporate income taxes – a practice known as “deferral”. As of 2015,US multinationals were estimated to have accumulated $2.6 trillion in undistributed earnings overseas (JCT,2015). In the UK, foreign earnings were often returned to multinationals’ parent companies as loans, butthe US taxes these transactions as “constructive dividends”.

7Worldwide taxation also makes countries less attractive as a place for locating headquarters and dis-advantages their firms when bidding for foreign assets against firms from territorial countries. To escapehome-country tax, multinational companies may “invert” - that is, reincorporate or seek acquisition by acompany headquartered in a territorial country.

5

This paper proceeds as follows. Section 2 provides some background information on the

worldwide and territorial tax systems implemented in the UK before and after 2009 and a

simple model to clarify the effects on profit shifting. Section 3 describes the data set and

gives some descriptive evidence. Section 4 presents the empirical strategy, and Section 5 our

results. Section 6 discusses the economic magnitude of our effects and the resulting policy

implications, including the effect of profit shifting on measured TFP. Section 7 concludes.

2 Institutional Background

2.1 Taxation of Foreign Earnings in the UK

Pre-2009 Worldwide Regime. Until 2009, the UK taxed the worldwide profits of firms

headquartered in the UK with deferral. Under this worldwide tax system, foreign profits of

UK multinationals were taxed at the domestic rate when repatriated as dividend payments

to the parent company in the UK. British firms obtained a credit for taxes paid in foreign

countries to avoid double taxation, and the amount of the foreign tax credit was capped at

the UK tax liability on those earnings. Thus, in principle, firms’ tax burden was determined

by the UK tax rate and did not depend on the country where the profits were generated (as

long as the foreign tax rate was lower than the UK tax rate). However, the UK deferred the

taxation of foreign income until the firm repatriated the profits.

Post-2009 Territorial Regime. In 2009, the UK moved to a territorial tax system, which

excludes foreign profits that a British firm receives from foreign subsidiaries from domestic

taxable income. Thus, all income earned abroad is exempted from UK profit taxes. Foreign

affiliates of UK multinationals pay profit taxes only in the country where production takes

place. The exemption is 100 percent for a wide range of foreign-source dividends, including

profits accumulated before the introduction of the new legislation. Unlike the recent move

in the US toward a more territorial system following the passage of the TCJA, the UK

reform did not impose any deemed tax on unrepatriated profits or any minimum tax without

6

deferral on profits earned abroad.8 The UK also did not accompany the reform with any

major changes to existing anti-tax avoidance rules.9

Expected impact on profit shifting. Under the worldwide tax system, shifting profit

abroad only defers the tax payment until the income is repatriated. Permanently avoiding

taxation in the home country is only possible if there is a tax holiday, either temporarily or

permanently by switching to a territorial tax system. Thus, incentives to shift profits abroad

should be stronger under a territorial tax system than under a worldwide tax system—a point

we show more rigorously in the short model below. This insight holds both for shifting profits

from home to a low-tax country, as well as for shifting profits from higher-tax to lower-tax

subsidiaries. Tax payments on reported profits in high-tax countries become less valuable

as they cannot offset additional UK taxes on repatriations from low-tax countries; and at

the same time, some portion of tax savings that were (re)captured by the UK treasury now

go to UK multinationals. All these considerations point to a higher reported profitability in

low-tax countries, and a possibly lower reported profitability in high-tax countries.

2.2 A Simple Model of Profit Shifting with Tax Regime Change

This section develops a simple model that shows how a shift from a worldwide tax system to

a territorial tax system increases incentives for profit shifting by multinational corporations

(MNCs). We start from a modified version of the model based on Hines and Rice (1994)

8The TCJA created a modified territorial tax system by introducing several provisions to reduce the extentof profit shifting under the new regime. The minimum tax on Global Intangible Low-Tax Income (GILTI)imposes a 10.5 percent minimum tax without deferral on foreign profits that exceed a firm’s “normal” return(defined in the law as 10 percent on the adjusted basis in tangible property held abroad). The new baseerosion and anti-abuse tax (BEAT) sets a minimum tax on otherwise deductible payments between a UScorporation and a foreign affiliate. To transition to the new system, the TCJA also created a new deemedrepatriation tax for previously accumulated and untaxed earnings of foreign affiliates of US firms, which is15.5 percent for cash and 8 percent for illiquid assets.

9The UK government revised its controlled-foreign-corporation (CFC) rule four years later (in 2013).At the time of the reform, all profits of a CFC (active or passive), were liable to UK taxes on a currentbasis. However, there were a series of exemptions from the CFC rules, including an exemption for activelytrading subsidiaries. The 2009 Finance Bill (which introduced dividend exemptions) only changed the CFCregime minimally in fear of hurting the UK’s ability to attract multinational firms. A new CFC regime wasenacted fully in January 2013. It focuses more narrowly on foreign profits artificially diverted from the UKby extending UK taxes to all passive income abroad. Thus, in contrast to the recent US reform, the UKtook a more stepwise approach by strengthening its CFC rules four years after introducing the territorialtax system.

7

and Grubert (1998), and additionally model repatriation taxes.

Basic setup. Consider an MNC with headquarter in country h and affiliates in n countries.

Let ρi present economic profits earned in each country by real factors located there. The

MNC engages in profit shifting and allocates additional πi of profits to affiliates in country i.

The cost of reporting a profit that is different from true profit rises quadratically in the

degree of profit shifting:

Ci =λ

2

(πi)2

ρi,

where λ > 0 captures the strength of tax enforcement.

While MNCs can decide where to report their book profits, they can only pay out di-

vidends to shareholders or reinvest at home if profits have been repatriated to the headquarter.

Under the worldwide system, firms pay dividend taxes τ repath on earnings repatriation. For-

eign taxes are credited against this tax, so that the total tax burden on foreign profits is

max(τ repath , τi).10 In a territorial system, there are no additional taxes on foreign repatri-

ations, so τ repath = 0. Assume that penalties are paid at home and are not tax deductible.

The after-tax profits for the MNC are given by:

Π = (1− τh)

[ρh −

n∑i=1

πi

]+

n∑i=1

(1−max(τ repath , τi)

)[ρi + πi]−

n∑i=1

λ

2

(πi)2

ρi. (1)

Optimal Profits Shifted. Taking the first-order condition with respect to πi, we derive:

πiρi

=−(1− τh) +

(1−max(τ repath , τi)

)λ

. (2)

The optimal amount of profits shifted, in proportion to economic profit, depends on the tax

differential between the two countries, as well as the strength of tax enforcement. The total

profits reported in country i, i.e. ρi + πi, relative to the true economic profits, is 1 + πiρi

.

Proposition 1 (Optimal Reported Profitability) Relative to the economic profits ρi,

10The tax burden on foreign profits (τ repath ) may be lower than the profit tax rate at home (τh) also undera worldwide tax system, e.g. because the repatriation tax has to be paid only later, when profits are indeedrepatriated.

8

(i) under a territorial tax system, the optimal reported profits increases in the tax differ-

ential τh − τi,

(ii) under a worldwide tax system, the optimal reported profits increases in the adjusted tax

differential τh −max(τ repath , τi). If τ repath = τh, no profit is shifted abroad.

Proof. Follows directly from equation (2).

Due to the repatriation tax, shifting profits abroad is less tax efficient for the MNC under

worldwide taxation. Therefore, the difference in the statutory tax rates distorts reported

profits less under worldwide taxation. A shift from a worldwide system with deferral to a

territorial system therefore increases the incentive for profit shifting to low-tax countries.

Proposition 2 (Tax Reform and Reported Profitability, Main Prediction) There is

more profit shifting under a territorial tax system than under a worldwide tax system:

πiρi

(τ repath > 0

)> πi

ρi

(τ repath = 0

).

Proof. Follows directly from equation (2).

3 Data and Descriptive Analysis

3.1 Data on UK Affiliates

Our empirical analysis uses affiliate-level information on the financial statements of mul-

tinational firms’ and their subsidiaries from Bureau van Dijk’s Amadeus database. This

database includes unconsolidated balance sheet information and ownership data for a large

sample of European firms. Our sample includes subsidiaries in all European countries (see

Appendix Table A.1), as well as information on those ultimate parent companies in the UK.

We include a subsidiary if at least 50 percent of its shares are (directly or indirectly) held

by the ultimate owner company.

The Amadeus database provides consistent information on a large number of firms in

Europe, but it does not include any of the small tax haven countries.11 Thus, our paper

11Using the corresponding worldwide database by Bureau van Dijk, Orbis, would not solve this problem:Tørsløv et al. (2018) show that tax haven affiliates are systematically missing in Orbis.

9

can only show that multinationals shift profits, but will underestimate the magnitude as

most tax avoidance occurs with tax haven counterparties (Davies et al., 2018). However, the

sample is well-suited to study the effect of profit shifting on an indicator of real economic

performance (in our case, TFP). Our results are also informative when one is interested in

the magnitude of profit shifting among affiliates set up for non-tax purposes.

The initial sample consists of multinational companies in the EU-27 countries from 2006

to 2012. We exclude companies from the finance industry and public utilities, as these

firms are highly regulated and face different tax avoidance incentives and opportunities.

Moreover, we drop observations if their industry classification is missing or if they have a

negative value for assets or employees. Our final sample is an unbalanced panel with 305,086

firm-year observations for 57,136 unique firms.

Our main sample of interest refers to firms that are active in a country with a lower tax

rate than the UK in all years in which we observe the firm. We refer to these observations

as “low-tax”. As both the UK and other countries changed their tax rates during the

sample period, we determine the “low-tax” status based on the individual affiliate. All

observations from Austria, Bulgaria, Czech Republic, Estonia, Finland, Greece, Hungary,

Ireland, Netherlands, Poland, Romania, Slovakia, Slovenia and Sweden are thus “low-tax”,

as well as some observations from Portugal (depending on how long they are in the sample).

In placebo tests, we study firms that are active in a country with a higher tax rate than the

UK in all years in which we observe it (“high-tax firms”). All observations from Belgium,

Denmark, France, Germany, Italy, Luxembourg, and Spain are considered “high-tax”.

Our main variables are firms’ earnings before tax (EBT), earnings before interest and tax

(EBIT), total assets, intangible assets, the number of workers, turnover and the investment

rate. The investment rate is defined as gross investment (proxied by the yearly change

in tangible and intangible assets, plus depreciation) divided by the sum of tangible and

intangible assets at the start of the year. Table 1 provides some descriptive statistics. Low-

tax firms in the sample have on average earnings before taxes of 870,979 Euro, total assets of

16 million Euro, sales of 19 million Euro and employ 97 workers. For all these variables, the

median lies substantially below the mean, indicating that the sample includes a substantial

number of smaller affiliates. We winsorize all firm-level ratio variables at the 1 percent and

10

99 percent level in the full sample (low-tax and high-tax combined).

Table 1. Summary Statistics: Affiliate-Level

Obs. Mean Median Std. Dev. Min Max

Low-Tax Observations

EBT (Euro) 155,076 870,979 31,907 4,802,172 -12,329,598 46,736,000EBIT (Euro) 155,076 889,759 49,091 4,149,189 -9,870,000 38,712,365EBT/Total assets 155,076 0.03 0.04 2.02 -363.96 427.00Total assets (Euro) 155,076 16,075,129 1,661,694 63,997,889 13,594 754,076,615Intangible assets (Euro) 152,169 264,648 94 2,005,064 0 28,187,285Workers 155,076 97 17 248 1 2,156Turnover (Euro) 155,076 19,468,500 2,032,027 67,916,729 5,148 675,585,141

High-Tax Observations

EBT (Euro) 150,010 1,622,421 144,026 7,044,888 -12,329,598 46,736,000EBIT (Euro) 150,010 1,514,149 178,666 5,816,367 -9,870,000 38,712,365EBT/Total assets 150,010 0.03 0.04 0.87 -288.40 76.71Total assets (Euro) 150,010 39,501,772 6,692,006 111,567,582 13,594 754,076,615Intangible assets (Euro) 148,473 1,039,982 12,589 4,074,373 0 28,187,285Workers 150,010 121 30 287 1 2,156Turnover (Euro) 150,010 41,859,669 9,015,245 103,072,763 5,148 675,585,141

Descriptive statistics for the main variables in the affiliate-level dataset used in the regressions. Low-tax(high-tax) observations are firms that are in countries which have a lower (higher) tax rate than the UK inall years in which we observe the firm. All ratio variables are winsorized at the 1 percent level. Data fromAmadeus for 2006-2012.

3.2 Data on UK Parents

We combine the unconsolidated data with information on consolidated financial statements

of UK-based corporations obtained from Thomson Reuters’ Datastream database. We merge

the two datasets based on the ultimate owner’s ISIN number. We are able to successfully

merge information on 595 ultimate owners (out of 1,239 parent companies). The matched

dataset is an unbalanced panel from 2005 to 2014 with a total of 3,587 firm-year observations

with non-missing information on the main regression variables. Due to missing information

in Datastream, there are more observations for more recent years. Table A.2 in the Appendix

summarizes how the observations are distributed over time.

Table 2 presents summary statistics for the consolidated firms. On average, the consol-

idated balance sheets show total assets of 2.3 billion Euro, and on average 19.6 percent are

11

intangible assets. The average consolidated firm has 10,599 employees and pre-tax income

of 228 million Euro. 43 percent of these firms have foreign assets, indicating that they are

multinational firms.

Table 2. Summary Statistics: Consolidated Firms

Obs. Mean Median Std. Dev. Min Max

Total assets (thous. Euro) 3,587 2,258,277 92,301 13,972,717 13 285,302,595Intangible assets (thous. Euro) 3,585 442,404 12,713 1,766,721 0 22,357,000Foreign assets (thous. Euro) 1,544 84,396 3,014 1,062,445 0 40,876,316Pre-tax income (thous. Euro) 3,587 227,639 3,800 1,831,377 -2,261,516 39,823,617EBIT (thous. Euro) 3,537 252,244 5,108 1,885,164 -2,112,087 40,265,783Total Debt (thous. Euro) 3,587 487,068 6,729 2,235,239 0 37,661,580Employees 3,587 10,599 526 46,760 1 648,254

Descriptive statistics for the main variables in the consolidated-firm-level dataset for regression ana-lysis. Data from Datastream for 2005-2014.

3.3 Descriptive Analysis

We begin our analysis by some simple comparisons of parent firms before and after the

UK’s change to a territorial tax system in 2009. If UK firms indeed shifted more profits

abroad after 2009, we should see that their effective foreign tax burden decreased after 2009.

Figure 1 Panel A plots the density of effective foreign tax rates of UK multinationals. Before

2009, a large number of UK multinationals had an effective foreign tax rate close to the UK

tax rate of 29%, which they would have had to pay anyways upon repatriation. After 2009,

the distribution of effective foreign tax rates is more even, with more firms having a lower

effective foreign tax rates (at around 20%). While gradual decline in the statutory CIT rates

(Figure 1 Panel B) may contribute to lower effective tax rate and shift their distribution

to the left, the diffusion of bunching around the pre-reform UK tax rate is more clearly

associated with the tax regime change.12

12One may wonder why UK multinationals did not shift all profits to tax havens after the reform. First,most profit shifting strategies require some real activities in the location to which profits are shifted. It istherefore favorable to shift profits also to low-tax countries in which the multinational already had affiliatesfor non-tax reasons. In addition, firms may want to use the retained earnings obtained via profit shifting tofinance investment in the subsidiaries.

12

Figure 1. Density of Parent Firms’ Effective Foreign Tax Rate Before and After 2009

A. Effective Foreign Tax Rates B. Statutory Foreign CIT Rates

Panel A: Kernel density of parent firms’ effective foreign tax rates pre-2009 (blueline) and post-2009 (dashed red line). Data from Datastream for 2005-2014. Panel B:Kernel density of statutory CIT rates in EU-28, pre-2009 (blue line) and post-2009(dashed red line). Vertical line marks the UK tax rate of 29%.

4 Empirical Strategy

4.1 Estimating Profit Shifting

We have highlighted in Section 2 that our identification relies on the variation in the incentive

for outward profit shifting arising from the UK’s switch to a territorial tax system. Naturally,

our main empirical strategy is a standard difference-in-differences (DiD) approach. We

estimate the difference in the reported profitability before and after 2009, comparing affiliates

of UK-based multinationals with affiliates of non-UK multinationals in the same country.

Following the discussions in Section 2, UK multinationals in low-tax countries have stronger

incentives for outward profit shifting after the reform, whereas profit shifting incentives for

non-UK-owned multinationals should remain the same. We thus expect a positive coefficient

on the DiD estimator on reported profitability in low-tax countries.

Formally, we test the following regression specification:

yikt = β1UK-Parenti × Postt + βxXikt + βzZkt + αi + dt + εikt, (3)

13

where yikt is a measure of profitability of subsidiary i in country k in year t. The key variable

of interest is the interaction term UK-Parenti×Postt, where UK-Parenti is a dummy indicator

that takes the value of one for a UK-owned multinational affiliate, and zero otherwise, and

Postt is a dummy indicator that takes the value of one for all years after 2009. To control

for time-invariant subsidiary and parent characteristics, we include a full set of firm-fixed

effects (αi). We also control for time-varying input factor proxies Xikt, namely the natural

logarithms of the number of employees and of fixed tangible assets.13 We include year

dummies (dt) to capture the effect of aggregate macroeconomic shocks that are common

to all companies. In our preferred specifications, we include time-varying country- and

industry-trends to control for such shocks in a more flexible way.

The firms in our sample have parent companies in different countries. We thus control for

some characteristics of these countries, which would, for example, capture aggregate shocks

(such as the global financial crisis) that hit parent countries to a differing extent. These

home-country control variables Zkt are the inflation rate, GDP per capita, the long-term

unemployment rate, and GDP growth. Finally, εikt is the error term.

Common Trends Our identifying assumption is that in the absence of the territorial tax

reform the control firms would have similar trends in reported profitability to the treated

firms. While we cannot directly test this assumption, we check whether reported profitability

trended similarly in the control and treated firms in the pre-reform period. To validate the

empirical design, we extend the DiD analysis to an event-time specification by estimating

the following model:

yikt =3∑

θ=−3

δt1[t = θ]× UK-Parenti + βxXikt + βzZkt + αi + dt + εikt, (4)

where 1[t = θ] is a series of year dummies that equal one when the territorial tax reform is θ

years away. Each coefficient δt measures the change in profitability for treated firms relative

to control firms in the θ-th year before or after the reform became effective in 2009. The

omitted time category is θ = −1, so that the estimated effects δt are relative to the period

13Given that the ownership information in our sample is time invariant, firm fixed effects also subsumehost-country fixed effects.

14

prior to the start of the reform. By focusing on the lead variables (i.e., the year dummies

leading up to the reform), this specification allows us to check whether there is any evidence

of pre-existing trends. In addition, the lag variables (i.e., the year dummies after the reform)

illustrate the potential dynamic effects of the reform.

Common Trends across Company Groups As a step further, we test whether changes

in the profitability at the affiliate level are indeed due to profit shifting or merely reflect

changes in the overall profitability of UK multinational groups after reform (e.g. because of

lower cost of capital due to the lower tax rate, or simply because they were more resilient

to the financial crisis). We use the consolidated data and compare UK multinationals with

domestic companies in the UK. We estimate the following equation:

yjt = γ1MNCj × Postt + γxXjt + ηj + dt + εjt, (5)

where y is the profitability of the consolidated firm j in year t. In this specification, the

main variable of interest is the interaction term MNCj × Postt, where MNCj is a dummy

variable that is one if the UK company group has foreign affiliates, and Postt is again a

dummy variable that is one for the years after 2009. We add consolidated-firm fixed effects

ηj to control for time-constant firm characteristics, and control for several time-varying firm

characteristics, Xjt, including the natural logarithms of workers and fixed assets, as well

as turnover growth. We also add year dummies dt to control for overall shocks to the UK

economy. With consolidated data capturing the overall profitability of the company group,

which should be invariant to inter-company profit shifting, we should not find a significant

coefficient for γ1 in this specification.

4.2 Key Dependent Variables

Pre-tax profitability The literature has used different measures of pre-tax reported prof-

itability, e.g. the ratio of earnings before tax to total assets (EBT/Assets, see e.g. Schwarz,

2009; Loretz and Mokkas, 2015) and the natural logarithm of earnings before tax (ln (EBT),

see e.g. Dharmapala and Riedel, 2013; Dischinger et al., 2014). While we report results from

15

both measures in Section 5, we use the former as our preferred measure of profitability. By

taking the logarithm of EBT, we would exclude all affiliates with negative and zero earn-

ings before tax from our regressions, even though zero profit (or loss making) may reflect

profit shifting in the most aggressive form. Shifting to a loss-making affiliate saves taxes

efficiently, as it does not induce any positive taxes in the host country (Simone et al., 2017;

Hopland et al., 2018). Therefore, excluding loss-making affiliates from the analysis may lead

to downward bias in our results.

While we defer discussing the implication of using different profitability measures to

Section 5.1, Figure 2 shows that as a simple feature of the data the distribution of pre-

tax profitability has large dispersion and is centred around zero. Around 31 percent of

observations in our data report negative or zero EBT. Despite the fact that accounting

profits may be a noisy measure of taxable profit, there is clear evidence of bunching at zero,

a phenomenon that has also been documented in other recent studies (Johannesen et al.,

2019; Bilicka, 2019; Koethenbuerger et al., 2019; Hopland et al., 2018). In both the low-

tax and the high-tax group, more than five percent of companies report a return to assets

between 0 and 0.01 percent.

Figure 2. Distribution of Pre-Tax Profitability

Distribution of pre-tax profitability (measured by EBT/total assets)in low-tax and high-tax countries. Data from Amadeus for 2006-2012.

16

Total Factor Productivity The primary measurement of firm productivity (i.e., TFP)

is based on the Levinsohn and Petrin (2003) approach. To illustrate, consider a standard

logarithmized Cobb–Douglas production function in the form of

lnY kit = β0 + βK lnKk

it + βL lnLkit + βM lnMkit + vit + εit, (6)

where Y kit , K

kit, L

kit, and Mk

it refer to firm i’s output (measured by total turnover), capital

stock, labor, and materials in industry k in year t, respectively, and ωit = β0+vit presents the

firm-level productivity. Traditionally, the simplest benchmark of TFP is the Solow residual

from the ordinary least square (OLS) regression for specification (6), which is usually run

industry-by-industry and based on firm-level data. That is,

ln TFPkit = lnY k

it − ln Y kit .

However, the OLS estimates of specification (6) are likely to be inconsistent and biased,

largely due to endogeneity of input choices and selection biases. The “endogeneity of inputs”

is caused by correlation between the level of inputs chosen and unobserved productivity

shocks (De Loecker, 2011). In addition, firms with low productivity are more likely to exit

from the market, leading to a selection bias in the sample.

A common measure to address these biases is proposed in Levinsohn and Petrin (2003)

(hereby the LP approach), which uses intermediate inputs (such as raw materials, electricity,

or fuels) rather than investment as a proxy for the unobserved productivity shocks. In light

of the relative advantages of both approaches, we employ the LP approach as the primary

measure of firms’ TFP, using costs of goods sold to measure intermediate inputs. To check

the robustness of the TFP measure, we also use a factor share approach that calculates the

parameters from cost share data in equation (6) for each firm. By imposing constant returns

to scale in equation (6), the factor share approach of the log TFP is:

ln TFPit = ln(YitLit

k

)− βk ln(Kit

Lit

k

). (7)

where the cost share of capital is calculated separately for each country–industry pair, as

17

the mean value of one minus the share of labor costs in value added for firms in that country

and industry.

5 Evidence on Profit Shifting

5.1 Baseline results

Table 3 presents the results of our baseline specification (3), starting by using EBT/Assets as

the dependent variable. In Column (1), we estimate a baseline regression by controlling only

for firm labor inputs. We find a positive and highly significant coefficient on the interaction

term of 0.019, indicating that UK multinationals indeed shifted more profits to low-tax

affiliates after the introduction of the territorial tax system. Column (2) further controls

for the scale of output by including the natural logarithm of firm turnover, which leads to a

slightly smaller coefficient on the interaction term. By conditioning on the scale of output,

the estimation in Column (2) excludes transfer mispricing via goods and services.

In Columns (3) and (4), we add time-varying host country and time-varying industry

trends, which also effectively controls for the effect of statuary corporate tax rate changes

in the host country on profit shifting. The DiD coefficients are almost unaffected, with an

estimated value of 0.021 without controlling for turnover, and of 0.017 when controlling for

turnover.14 Our preferred specification is Column (3), which allows for profit shifting through

a broad range of channels while controlling for country-specific and industry-specific time-

varying shocks. Lastly, in Column (5), we differentiate between wholly-owned and majority-

owned subsidiaries. It is more attractive to shift profits to a wholly-owned subsidiary, as

otherwise minority shareholders profit from the profit shifting activities as well. Correspond-

ingly, we find that profit is only shifted to the wholly-owned subsidiaries.

We next focus on profitable affiliates. To do so, we now use the natural logarithm of earn-

ings before tax (ln(EBT )) as the dependent variable. This specification allows interpreting

the estimated coefficient as a tax semi-elasticity and facilitates comparison with the wider

literature on profit shifting. Column (6) reports the results, showing that among affiliates

14We now also include controls for several characteristics of the MNC parent’s home country.

18

Table 3. Profit Shifting after Territorial: Baseline Results

Sample: Low-Tax High-Tax

Dependent variable: EBT/Assets ln (EBT) EBT/Assets

(1) (2) (3) (4) (5) (6) (7)

UK-Parent × Post 0.019*** 0.017*** 0.021*** 0.017** 0.111** 0.002(0.007) (0.006) (0.007) (0.007) (0.045) (0.006)

UK-Parent × Post 0.031***× Wholly-Owned (0.009)UK-Parent × Post 0.011× Majority-Owned (0.009)ln (Workers) 0.016*** -0.035*** 0.016*** -0.035*** 0.016*** 0.309*** 0.009***

(0.002) (0.002) (0.002) (0.002) (0.002) (0.014) (0.002)ln (Turnover) 0.071*** 0.070***

(0.002) (0.002)ln (Fixed Assets) 0.092***

(0.007)Year FEs Y Y - - - - -Affiliate FEs Y Y Y Y Y Y YIndustry-Year FEs - - Y Y Y Y YHost Country-Year FEs - - Y Y Y Y Y

N 155,076 155,076 155,076 155,076 155,076 97,357 150,010R2 0.487 0.521 0.495 0.529 0.495 0.851 0.591

Data from Amadeus for 2006-2012. Specification (3) to (7) include the following characteristics of themultionational firm’s home country: inflation, GDP per capita, long-run unemployment rate, and GDPgrowth. Standard errors in parentheses are clustered by parent. ***, ** and * indicate significance at the1%, 5% and 10% levels.

19

with positive pre-tax earnings, on average their EBT increases by 11 percent following the

territorial tax reform. In comparison, the estimated coefficient in column (3) suggests that

on average the pre-tax EBT (scaled by total assets, with a pre-reform mean of 0.06 for UK

firms) increases by 34 percent for the full sample, an effect that is three time larger than

when only focusing on the profitable affiliates.

Placebo tests. The validity of our main results in Table 3 relies on the assumption that

there were no differential trends from the treated and control firms in the pre-reform period.

To verify this assumption, we conduct an event study by estimating specification (4) repla-

cing the reform dummy in Column (3) of Table 3 with a series of year dummies for a window

of three years around the territorial tax reform. Figure 3 plots the estimated coefficients on

these dummies and shows the 90 percent level confidence intervals. These are interpreted

as the differential changes in pre-tax profits reported by UK-owned multinational affiliates,

relative to the non-UK affiliates, as compared to the last year prior to the implementation

of the territorial reform. As the graph shows, the parallel pre-trend on pre-tax profitability

between the treated and control firms is satisfied, as the coefficient estimates are close to

zero prior to the reform. Moreover, positive and statistically significant coefficients after the

year 2009 indicate that the territorial tax reform induced the affiliates of UK multinationals

to report more profits. The gradual increase in the profitability after the reform is consist-

ent with somewhat delayed responses due to adjustment costs associated with the different

channels of profit shifting.15

As an alternative test, we run the regression in Column (3) of Table 3 for high-tax

affiliates. The results are reported in Column (7) of the same table. The coefficient of the

interaction term is statistically insignificant and close to zero. This zero result is relatively

precisely estimated; its standard error is similar to the other regressions. We can interpret

this test as a placebo test, as there is no incentive to shift profits from the UK to high-tax

countries.

This test also shows that the additional profits shifted to low-tax countries do not origin-

ate in affiliates in high-tax countries (which would be consistent with a negative coefficient).

15Some caution need to be taken with this interpretation, since the post-reform coefficient estimates arenot statistically different from each other.

20

Figure 3. Profit Shifting after Territorial: Placebo Results

Regression coefficients from estimating eq. (4), replacing the reform dummy withyear dummies for the three years before and after the reform. Bars depict 90 percentconfidence intervals. Data from Amadeus for 2006-2012.

Thus, multinationals seem to increase profit shifting from the UK headquarter to the low-

taxed affiliates after the reform.

Tests with consolidated data. To investigate this question further, we now turn to the

results with the consolidated data and compare UK multinationals to UK domestic firms.

These tests also allow us to rule out that the increase in profit shifting to low-tax countries

is not an artifact of higher overall profitability of UK multinationals after the territorial tax

reform.

Table 4 presents these results. In column (1), we estimate eq. (5) without control vari-

ables. We find an insignificant, but sizable coefficient of 0.184. This potential effect could

arise because firms increased investment after the reform (see Liu, 2019). Indeed, when we

include control variables in column (2), the estimated coefficient becomes much closer to

zero. This finding remains unchanged when using the ratio of earnings before taxes relative

to total assets as the dependent variable in column (3). Thus, the profitability of the consol-

21

idated accounts of UK MNCs remains approximately constant after the reform, indicating

that the effects found on earnings before taxes on the affiliate level indeed arise because of

profit shifting. As the profitability of high-tax affiliates does not change, the shifted profits

must originate in the UK headquarter.

Table 4. DiD Tests with Consolidated Data

Sample: Consolidated Accounts

Dependent variable: ln(EBT) EBT/Assets

(1) (2) (3)

MNC × Post 0.184 0.068 0.003(0.118) (0.103) (0.014)

ln (Fixed Assets) 0.542***(0.083)

ln (Workers) 0.113 -0.011(0.078) (0.020)

Sales growth rate 0.002(0.001)

Year FEs Y Y YFirm FEs Y Y YN 2,431 2,431 3,587R2 0.04 0.12 0.01

The dependent variable are the natural logarithm of earnings beforetaxes (in columns 1 and 2), and earnings before taxes relative to totalassets (in column 3). Data from Datastream for 2005-2014. Robuststandard errors in parentheses. ***, ** and * indicate significance at the1%, 5% and 10% levels.

Robustness checks. To test for the robustness of the basic results, we conduct sensitivity

analysis along three dimensions. First, we define “low-tax” and “high-tax” countries using

an alternative approach, where a country is considered as low-tax (high-tax) if its statutory

corporate tax rate is always below (above) the UK rate during 2005-2012. Table 5 Column

(1) reports the results using this alternative definition of low-tax countries. Results are

basically unchanged compared to the baseline findings. Second, we create a subsample

of matched firms to address the concern that companies in the treated UK and control

22

affiliates may not have similar observable characteristics, and that these differences may

explain different trends in reported profitability over time. The sample is matched from

a Mahalanobis distance matching procedure based on pre-reform firm-level turnover, fixed

assets, and employment. Column (2) replicates the DiD analysis on this matched sample

in low-tax countries. The resulting estimate has a wider confidence interval due to fewer

observations, but nevertheless, remains positive and significant at the 5% level. Column (3)

expands the low-tax sample to include observations in Switzerland, which features low tax

rate and strong privacy laws and is outside EU27. The results remain very similar. Panel

(B) of Table 5 reports the results concerning the high-tax countries, where the estimated

impact of the reform remains small and insignificant with alternative definition of high-tax

countries, a matched sample based on similar turnover, fixed assets, and employment, and

a smaller sample excluding observations from Luxembourg (which has a high nominal tax

rate, but is often considered a tax haven due to numerous special tax rules).

5.2 Heterogeneity Analysis

Recent literature has uncovered effect heterogeneity in the profit shifting behavior of mul-

tinational entities. For example, profit shifting is concentrated in large multinationals (Wier

and Reynolds, 2018), is stronger in R&D intensive firms (Liu et al., 2019) and to countries

that impose low or no taxes on corporate profits (Dowd et al., 2017; Davies et al., 2018;

Tørsløv et al., 2018). To explore the extent to which the impact of territorial tax reform

may vary across these characteristics, we divide firms in the low-tax sample into quintiles

(for each firm characteristic), and estimate the effect of the tax reform by interacting the

policy variable with the quintile indicators:

yikt =5∑j=1

βQuintilejUK-Parenti×Postt×1{i ∈ Quintilej}+βxXikt+βzZkt+αi+dt+εikt, (8)

where 1{i ∈ Quintilej} is the jth quintile indicator defined above, and all other variables

are as previously defined. Figure 4 summarizes the main results concerning the effect het-

erogeneity of the territorial tax reform, reporting the coefficient estimates βQuintile and the

95 percent confidence interval across each firm or country characteristic.

23

Table 5. Profit Shifting after Territorial: Robustness Results

Panel A. Low-Tax Countries

Alternative Definition Matched Addingof Low-Tax Countries Sample Switzerland

(1) (2) (3)

UK-Parent × Post 0.020*** 0.029** 0.019***(0.007) (0.014) (0.007)

N 165,360 11,658 185,878R2 0.49 0.51 0.49

Panel B. High-Tax Countries

Alternative Definition Matched Excludingof High-Tax Countries Sample Luxembourg

(1) (2) (3)

UK-Parent × Post 0.002 0.004 0.003(0.006) (0.009) (0.006)

N 149,970 19,015 149,619R2 0.59 0.60 0.61

The dependent variable is earnings before taxes divided by total assets. Data fromAmadeus for 2006-2012. Standard errors in parentheses are clustered by parent.***, ** and * indicate significance at the 1%, 5% and 10% levels.

24

Firm size Panel A of Figure 4 plots the results across firm sizes measured by fixed assets.

It shows that only medium-to-large UK affiliates in the upper quintiles of the fixed-asset

distribution significantly increased their reported profitability in response to the 2009 reform.

Panel B shows a stronger pattern across the size measured by employment. The evidence

shows the effect of the reform on reported profits is concentrated in larger firms.

Intensity of intangibles Next, we test whether firms with intangible assets indeed have

larger opportunities to shift profits to low-tax affiliates. Panel C of Figure 4 plots the results

across the ratio of intangible assets. The evidence shows that the effect of the reform on

reported profits is larger in firms with relatively high levels of intangibles.

Level of host-country tax rate Lastly, we test whether the effect differs in the level of

statutory tax rate in the host country. Panel D of Figure 4 plots the results in countries with

relatively low, medium, and high tax rates within the low-tax country sample. Surprising,

there is no evidence that profit shifting increases with the tax rate differential.16 As our

sample includes a large number of loss-making affiliates, the statutory tax rate does not

capture the underlying profit shifting incentives well (see Simone et al., 2017; Hopland et al.,

2018; Gamm et al., 2018). Moreover, there may well exist tax incentives for profit shifting

in a host country, including from company-specific advance pricing agreements and country-

specific double tax treaty networks that are not reflected in the statutory tax rate. Differences

in enforcement of anti-avoidance regulations across countries can be another possible reason

why the tax differential effect is non-monotonic.

16When imposing restriction that shifting increases linearly with the tax rate differential, the coefficienton the interaction term UK-parenti × Postt × δτit is also insignificant.

25

Figure 4. Effect Heterogeneity in Low-Tax Countries

A. Fixed Asset B. Employment

C. Intangible Asset D. Tax Rate

Estimated coefficients from specification (8) for quintiles, including 95% confidence intervalsacross firm size (measured by fixed assets in Panel A and employment in Panel B), intensity ofintangibles (Panel C), and host-country tax rate (Panel D). Data from Amadeus 2006-2012.

26

6 Economic Magnitude and Policy Implication

6.1 Economic magnitude

Our preferred results based on Column 3 of Table 3 suggest that there is a 2.1 percentage-

point increase in the overall pre-tax profitability of UK affiliates. Among profitable firms,

the reform increases the reported profitability by 11 percent.

To put these results in perspective, we use the consensus semi-elasticity of reported profits

to corporate tax rate differential of 1.5 in a recent meta-analysis (Beer et al., 2019), which

suggests that for every one percentage point lower corporate tax rate (everything else equal),

there is a 1.5 percent increase in the pre-tax profit reported by multinational affiliates. The

benchmark specification of the meta-analysis uses the logarithm of reported profit before

taxation as dependent variable, which is comparable to our results focusing on the level of

reported profitability in response to the territorial tax reform (Column 7, Table 3). On

average we find that for profit-making UK affiliates, there is an eleven percent increase in

their pre-tax earnings following the reform. The extent of increase in pre-tax profits is thus

equivalent to a seven percentage point reduction in the statutory corporate income tax rate

in the UK (= 11/1.5).

In terms of economic magnitude, overall the reported pre-tax profitability for UK affiliates

in low-tax countries in our sample increases by 2.1 percentage point following the reform.

Evaluated at the mean asset value for UK affiliates in the low-tax countries, the result implies

an increase of EUR 0.06 million for the median firm and EUR 0.65 million on average.

6.2 The effect of profit shifting on measured TFP

By altering the allocation of reported profits across countries, profit shifting by multinational

firms not only changes the distribution of tax revenue among the countries involved. It also

has implications on real economies in these countries as it affects the investment, employment

and the overall scale of production by multinational companies (see, e.g. de Mooij and

Liu, 2020; Suarez Serrato, 2018; Overesch, 2009). The impact of profit shifting on firm-

level behavior can also manifest and lead to distortions in the measurement of aggregate

27

statistics, for example in domestic GDP, productivity or the trade balance (Tørsløv et al.,

2018; Guvenen et al., 2017).

At the firm level, how profit shifting affects measured productivity depends on the specific

channels through which profit shifting takes place. The manipulation of transfer prices may

inflate reported productivity in low-tax countries as sales would be over-recorded and costs

of intermediate inputs would be under-recorded.17 Tax-motivated relocation of intellectual

property would affect measured productivity in low-tax countries by inflating both turnover

(via fee payments for the use of the intellectual property) as well as capital (the intellectual

property itself). On the other hand, profit shifting via internal financing should not affect

productivity measures as long as it does not change the level of reported operating profit or

turnover.

As Heckemeyer and Overesch (2017) show that firms’ non-financial inter-company trans-

actions (including tax-motivated transfer pricing and licensing) are the dominant profit-

shifting strategies, accounting for about 70-80% of the response to taxation. Thus, we

expect that stronger profit shifting following the territorial tax reform exacerbated the mis-

measurement of the productivity of UK multinationals in foreign countries.18

To test this hypothesis, we use two alternative measures of TFP, including one based on

the factor share approach and one estimated following the LP algorithm. The measurement

of output in calculating the TFP is not corrected for profit shifting, thus may be influenced

by the changing incentives of profit shifting after the territorial reform. We run a similar

difference-in-differences regression based on eq. (3), where the dependent variable is the

natural logarithm of TFP. Consistent with increased profit shifting into low-tax countries by

UK affiliates, we expect a corresponding increase in the measured TFP for UK affiliates and

hence a positive coefficient for β1.

Table 6 summarizes the regression results. Columns (1) and (2) focus on the low-tax

countries, respectively using the factor-share TFP and LP TFP measure as the dependent

variable. Both columns report a positive and significant coefficient on the interaction term,

17The reverse is true for high-tax countries.18In the longer term, if low-tax affiliates use inbound shifted profits to invest in R&D, profit shifting may

also lead to an increase in the real TFP. However, given the considerable time lag and uncertainty betweeninnovation input and output, we would expect the real impact of profit shifting on TFP to show up withsignificant delay rather than immediate following the territorial reform.

28

suggesting an increase in the measured TFP of between 5 and 9 percent for low-tax UK

affiliates after the territorial tax system. The results are in line with our baseline findings

on profit shifting. Suppose that 70-80 percent of profit shifting is through non-financial

transactions and captured in firm output variables that are typically used to calculate TFP,

then we would expect that on average, the measured TFP would increase by about 7-8

percent (given that the overall increase in the level of pre-tax profits is 11 percent).

Table 6. Profit Shifting after Territorial: Effects on TFP

Sample: Low-Tax High-Tax

Dependent variable: ln(TFPFS) ln(TFPLP ) ln(TFPFS) ln(TFPLP )

(1) (2) (3) (4)

UK-Parent × Post 0.093** 0.049* 0.003 0.005(0.044) (0.029) (0.021) (0.013)

Affiliate FEs Y Y Y YIndustry-Year FEs Y Y Y YHost Country-Year FEs Y Y Y Y

N 65,696 66,072 90,681 92,431R2 0.989 0.735 0.989 0.773

The dependent variable is natural logarithm of total factor productivity (TFP). Data fromAmadeus for 2006-2012. Standard errors in parentheses are clustered by firm. ***, ** and* indicate significance at the 1%, 5% and 10% levels.

As a placebo test, Columns (3)-(4) report the results for high-tax countries. The estim-

ated effect of the territorial tax reform is much smaller and insignificant in both specifica-

tions.

7 Conclusion

In this paper, we have used the introduction of a territorial tax system in the UK to study

the difference in profit shifting under a worldwide and a territorial tax system. Using a

difference-in-differences research design comparing the foreign subsidiaries of UK parents

with subsidiaries from other multinationals in the same country, we found that UK mul-

29

tinationals shifted significantly more profits to low-tax countries after the UK introduced

the territorial tax system. In placebo tests considering high-tax countries, we have found

no effect. In further analyses, our paper confirmed that firms with high shares of intangible

assets are more responsive to profit shifting incentives. We also find that increased profit

shifting following the territorial tax reform has a significant and positive impact on measured

TFP of UK affiliates in low-tax countries.

Our results have important implications for other countries switching to a territorial

tax system. In particular, the US introduced a territorial tax system as part of its tax

reform in December 2017. It is likely that the US will also see an increase in profit shifting,

counteracted by the newly-introduced anti-tax avoidance rules. Our results point out that

this response may be somewhat delayed, as the UK multinationals responded more strongly

about two years after the reform. Our results also highlight the interactive effects of profit

shifting and other economic indicators such as firm TFP, and caution the potential bias in

measurement of these variables due to profit shifting.

30

References

Arena, M. P., Kutner, G. W., 2015. Territorial Tax System Reform and Corporate FinancialPolicies. Review of Financial Studies 28, 2250–2280.

Azemar, C., Dharmapala, D., 2019. Tax Sparing Agreements, Territorial Tax Reforms, andForeign Direct Investment. Journal of Public Economics 169, 89–108.

Barrios, S., Huizinga, H., Laeven, L., Nicodeme, G., 2012. International taxation and mul-tinational firm location decisions. Journal of Public Economics 96, 946–958.

Becker, J., Fuest, C., 2011. The Taxation of Foreign Profits - the Old View, the New Viewand a Pragmatic View. Intereconomics 46, 92–97.

Beer, S., De Mooij, R. A., Liu, L., 2019. International Corporate Tax Avoidance: A reviewof the channels, effect sizes and blindspots. Journal of Economic Surveys (forthc.).

Bilicka, K. A., 2019. Comparing UK Tax Returns of Foreign Multinationals to MatchedDomestic Firms. American Economic Review 109, 2921–53.

Buettner, T., Wamser, G., 2013. Internal Debt and Multinationals’ Profit Shifting - EmpiricalEvidence from Firm-Level Panel Data. National Tax Journal 66, 63–96.

Clausing, K. A., 2003. Tax-Motivated Transfer Pricing and US Intrafirm Trade Prices.Journal of Public Economics 87, 2207–2223.

Clausing, K. A., 2009. Multinational Firm Tax Avoidance and Tax Policy. National TaxJournal 62, 703–725.

Clausing, K. A., Shaviro, D., 2011. A Burden-Neutral Shift from Foreign Tax Creditabilityto Deductibility?. Tax Law Review 64, 431–452.

Davies, R. B., Martin, J., Parenti, M., Toubal, F., 2018. Knocking on Tax Haven’s Door:Multinational Firms and Transfer Pricing. The Review of Economics and Statistics 100,120–134.

De Loecker, J., 2011. Product Differentiation, Multiproduct Firms, and Estimating the Im-pact of Trade Liberalization on Productivity. Econometrica 79, 1407–1451.

Desai, M. A., Foley, C. F., Hines, J., 2006. The Demand for Tax Haven Operations. Journalof Public Economics 90, 513–531.

Desai, M. A., Hines, J. R., 2003. Evaluating International Tax Reform. National Tax Journal56, 487–502.

Desai, M. A., Hines, J. R., 2004. Old Rules and New Realities: Corporate Tax Policy in aGlobal Setting. National Tax Journal 57, 937–960.

Devereux, M. P., Fuest, C., Lockwood, B., 2015. The Taxation of Foreign Profits: A UnifiedView. Journal of Public Economics 125, 83 – 97.

31

Dharmapala, D., 2014. What Do We Know About Base Erosion and Profit Shifting? AReview of the Empirical Literature. Fiscal Studies 35, 421–448.

Dharmapala, D., Foley, C. F., Forbes, K. J., 2011. Watch What I Do, Not What I Say:The Unintended Consequences of the Homeland Investment Act. Journal of Finance 66,753–787.

Dharmapala, D., Riedel, N., 2013. Earnings Shocks and Tax-Motivated Income-Shifting:Evidence from European Multinationals. Journal of Public Economics 97, 95–107.

Dischinger, M., Knoll, B., Riedel, N., 2014. The Role of Headquarters in Multinational ProfitShifting Strategies. International Tax and Public Finance 21, 248–271.

Dowd, T., Landefeld, P., Moore, A., 2017. Profit Shifting of U.S. Multinationals. Journal ofPublic Economics 148, 1 – 13.

Egger, P., Merlo, V., Ruf, M., Wamser, G., 2015. Consequences of the New UK Tax Exemp-tion System: Evidence from Micro-level Data. Economic Journal 125, 1764–1789.

Feld, L. P., Ruf, M., Scheuering, U., Schreiber, U., Voget, J., 2016. Repatriation Taxes andOutbound M&As. Journal of Public Economics 139, 13–27.

Flaaen, A., 2017. The Role of Transfer Prices in Profit-Shifting by U.S. Multinational Firms:Evidence from the 2004 Homeland Investment Act. Board of Governors of the FederalReserve System Finance and Economics Discussion Series 2017-055.

Gamm, M., Heckemeyer, J. H., Koch, R., 2018. Profit Shifting and the Marginal Tax Rate:What Determines the Shift-to-Loss Effect?. Available at ssrn.com/abstract=3172293.

Grubert, H., 1998. Taxes and the Division of Foreign Operating Income among Royalties,Interest, Dividends and Retained Earnings. Journal of Public Economics 68, 269–290.

Guvenen, F., Mataloni, J., Raymond J, Rassier, D. G., Ruhl, K. J., 2017. Offshore ProfitShifting and Domestic Productivity Measurement. NBER Working Paper 23324.

Hasegawa, M., Kiyota, K., 2017. The Effect of Moving to a Territorial Tax System on ProfitRepatriation: Evidence from Japan. Journal of Public Economics 153, 92–110.

Heckemeyer, J. H., Overesch, M., 2017. Multinationals’ Profit Response to Tax Differentials:Effect Size and Shifting Channels. Canadian Journal of Economics 50, p. 965–994.

Hines, B. J. R., 1996. Altered States : Taxes and the Location of Foreign Direct Investmentin America. American Economic Review 86, 1076–1094.

Hines, J. R., Rice, E. M., 1994. Fiscal Paradise: Foreign Tax Havens and American Business.Quarterly Journal of Economics 109, 149–182.

Hopland, A. O., Lisowsky, P., Mardan, M., Schindler, D., 2018. Flexibility in Income Shiftingunder Losses. The Accounting Review 93, 163–183.

32

Huizinga, H., Laeven, L., 2008. International Profit Shifting within Multinationals: A Multi-Country Perspective. Journal of Public Economics 92, 1164–1182.

IMF, 2019. Corporate Taxation in the Global Economy. International Monetary Fund IMFPolicy Paper 19/007.

Johannesen, N., Tørsløv, T., Wier, L., 2019. Are Less Developed Countries More Exposedto Multinational Tax Avoidance? Method and Evidence from Micro-Data. World BankEconomic Review 00, 1–20.

Koethenbuerger, M., Mardan, M., Stimmelmayr, M., 2019. Profit Shifting and InvestmentEffects: The Implications of Zero-Taxable Profits. Journal of Public Economics 173, 96–112.

Levinsohn, J., Petrin, A., 2003. Estimating Production Functions Using Inputs to Controlfor Unobservables. Review of Economic Studies 70, 317–341.

Liu, L., 2019. Where Does Multinational Investment Go with Territorial Taxation? Evidencefrom the UK. American Economic Journal: Economic Policy (forthc.).

Liu, L., Schmidt-Eisenlohr, T., Guo, D., 2019. Transfer Pricing and Tax Avoidance: Evidencefrom Linked Tax-Trade Statistics in the UK. Review of Economics and Statistics (forthc.).

Loretz, S., Mokkas, S., 2015. Evidence for Profit Shifting with Tax-sensitive Capital Stocks.FinanzArchiv 71, 1–36.

Maffini, G., 2012. Territoriality, Worldwide Principle, and Competitiveness of Multinationals:A Firm-Level Analysis of Tax Burdens.

Markle, K., 2016. A Comparison of the Tax-Motivated Income Shifting of Multinationals inTerritorial and Worldwide Countries. Contemporary Accounting Research 33, 7–43.

de Mooij, R., Liu, L., 2020. At A Cost: the Real Effects of Transfer Pricing Regulations.IMF Economic Review (forthc.).

Overesch, M., 2009. The Effects of Multinationals’ Profit Shifting Activities on Real Invest-ments. National Tax Journal 62, 5–23.

Richman, P. B., 1963. Taxation of Foreign Investment Income–An Economic Analysis. Bal-timore, Johns Hopkins Press.

Schwarz, P., 2009. Tax-Avoidance Strategies of American Multinationals: An EmpiricalAnalysis. Managerial and Decision Economics 30, 539–549.

Simone, L. D., Klassen, K. J., Seidman, J. K., 2017. Unprofitable Affiliates and IncomeShifting Behavior. The Accounting Review 92, 113–136.

Suarez Serrato, J. C., 2018. Unintended Consequences of Eliminating Tax Havens. NBERWorking Paper 24850.

33

Tørsløv, T. R., Wier, L. S., Zucman, G., 2018. The Missing Profits of Nations. NBERWorking Paper 24701.

Voget, J., 2011. Relocation of Headquarters and International Taxation. Journal of PublicEconomics 95, 1067–1081.

Weichenrieder, A. J., 2009. Profit Shifting in the EU: Evidence from Germany. InternationalTax and Public Finance 16, 281–297.

Wier, L., Reynolds, H., 2018. Big and ‘unprofitable’. UNU-WIDER Working Paper 111/2018.

34

8 Appendix

Table A.1. Observations by Country

Low-Tax High-Tax Total

Austria 3,874 0 3,874Belgium 0 22,274 22,274Bulgaria 7,921 0 7,921Czech Republic 31,520 0 31,520Denmark 0 20,090 20,090Estonia 8,166 0 8,166Finland 5,256 0 5,256France 0 36,460 36,460Germany 4,736 0 4,736Hungary 4,844 0 4,844Ireland 3,073 0 3,073Italy 0 34,729 34,729Luxembourg 0 391 391Netherlands 3,756 0 3,756Poland 19,763 0 19,763Portugal 461 40 501Romania 57,103 0 57,103Spain 0 36,026 36,026Sweden 820 0 820Slovenia 3,783 0 3,783

Total 155,076 150,010 305,086

Observations by country in the affiliate-level dataset.Low-tax (high-tax) observations are firms that are incountries with a lower (higher) tax rate than the UKin all years in which we observe the firm. Data fromAmadeus for 2006-2012.

35

Table A.2. Observations by Year (Consolidated Data)

# Obs. % of sample

2005 93 2.592006 95 2.652007 175 4.882008 197 5.492009 503 14.022010 523 14.582011 520 14.502012 510 14.222013 487 13.582014 484 13.49

Total 3,587 100

Observations by year in the consolid-ated data. Data from Datastream for2005-2014.

36